biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix][V1] Fix allowed_token_ids for v1 Sampler (#14169)

Browse Source 
Signed-off-by: Lu Fang <lufang@fb.com>
This commit is contained in:
Lu Fang
2025-03-05 00:49:44 -08:00
committed by GitHub
parent ec79b67c77
commit 8d6cd32b7b
2 changed files with 12 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/v1/engine/processor.py
View File

@@ -92,10 +92,12 @@ class Processor:
return
if params.allowed_token_ids is None:
return
if not all(0 <= tid < self.model_config.vocab_size
for tid in params.allowed_token_ids):
if not params.allowed_token_ids:
raise ValueError("allowed_token_ids is not None and empty!")
vocab_size = self.model_config.get_vocab_size()
if not all(0 <= tid < vocab_size for tid in params.allowed_token_ids):
raise ValueError(
"allowed_token_ids contains out-of-vocab token id")
"allowed_token_ids contains out-of-vocab token id!")
def process_inputs(
self,